ybelkada commited on
Commit
cfd869e
1 Parent(s): 15025be

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +4 -3
README.md CHANGED
@@ -99,7 +99,7 @@ license: apache-2.0
99
 
100
  # TL;DR
101
 
102
- If you already know T5, FLAN-T5 is just better at everything. For the same number of parameters, these models have been fine-tuned on more than 1000 additional tasks covering also more languages.
103
  As mentioned in the first few lines of the abstract :
104
  > Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints,1 which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models.
105
 
@@ -153,7 +153,7 @@ print(tokenizer.decode(outputs[0]))
153
  <summary> Click to expand </summary>
154
 
155
  ```python
156
-
157
  from transformers import T5Tokenizer, T5ForConditionalGeneration
158
 
159
  tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-large")
@@ -176,6 +176,7 @@ print(tokenizer.decode(outputs[0]))
176
  <summary> Click to expand </summary>
177
 
178
  ```python
 
179
  import torch
180
  from transformers import T5Tokenizer, T5ForConditionalGeneration
181
 
@@ -197,7 +198,7 @@ print(tokenizer.decode(outputs[0]))
197
  <summary> Click to expand </summary>
198
 
199
  ```python
200
- # pip install bitsandbytes
201
  from transformers import T5Tokenizer, T5ForConditionalGeneration
202
 
203
  tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-large")
 
99
 
100
  # TL;DR
101
 
102
+ If you already know T5, FLAN-T5 is just better at everything. For the same number of parameters, these models have been fine-tuned on more than 1000 additional tasks covering also more languages.
103
  As mentioned in the first few lines of the abstract :
104
  > Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints,1 which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models.
105
 
 
153
  <summary> Click to expand </summary>
154
 
155
  ```python
156
+ # pip install accelerate
157
  from transformers import T5Tokenizer, T5ForConditionalGeneration
158
 
159
  tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-large")
 
176
  <summary> Click to expand </summary>
177
 
178
  ```python
179
+ # pip install accelerate
180
  import torch
181
  from transformers import T5Tokenizer, T5ForConditionalGeneration
182
 
 
198
  <summary> Click to expand </summary>
199
 
200
  ```python
201
+ # pip install bitsandbytes accelerate
202
  from transformers import T5Tokenizer, T5ForConditionalGeneration
203
 
204
  tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-large")